Analysis of Machine Translation and Post-Translation Editing Ability Using Semantic Information Entropy Technology

Area of Science:

• Computational Linguistics
• Artificial Intelligence
• Natural Language Processing

Background:

• Machine translation (MT) faces challenges with large-scale corpora.
• Current semantic feature extraction methods have significant flaws.
• Improving MT accuracy and efficiency is crucial for global communication.

Purpose of the Study:

• To investigate AI-driven machine translation and post-editing capabilities.
• To develop an automatic post-translation editing system for MT.
• To enhance semantic feature extraction and reduce common MT errors.

Main Methods:

• Examined target language grammar rules.
• Projected semantic information entropy and derived the semantic Gaussian marginal rectangular window function.
• Integrated semantic correlation factors into text information entropy and information gain to obtain nonlinear spectral properties of adaptive matching semantics.

Main Results:

• Achieved word translation accuracy of 95.27%.
• Achieved segmentation accuracy of 93.12%.
• Effectively corrected flaws in conventional semantic feature extraction.

Conclusions:

• The proposed automatic post-translation editing system significantly speeds up MT and enhances translation quality.
• The approach offers accurate and trustworthy language translation.
• This research provides a valuable resource for future MT studies.